Token-level hallucination detection
Benchmarks
DatasetSOTA methodMetricTrendResultsLast updated
88Precision
27
Apr 7, 2026
76.9AP
18
Jul 1, 2026
59.4AP (Token-level)
18
Jul 1, 2026
8.42S_incor
14
May 13, 2026
6.46S_incor
14
May 13, 2026
9.29S_incor Score
14
May 13, 2026
9.86S_incor
14
May 13, 2026
93.33AUROC
7
Apr 20, 2026
95.6AUROC
7
Apr 20, 2026
66Object Precision
7
Feb 26, 2026
94Object Precision
7
Feb 26, 2026
61Object Precision
7
Feb 26, 2026
89.47AUROC
5
May 13, 2026
87.39AUROC
5
May 13, 2026
77.03AUROC
5
Apr 20, 2026